Anthropic Overhauls Subscription Model, Cuts Third-Party Agent Support Amid Pentagon Fallout

The policy change, effective Saturday at 12 p.m. PT, requires users to purchase discounted usage bundles or obtain a separate API key to integrate third-party tools.

Changes to Third-Party Tool Access

Anthropic has ended direct support for third-party AI agent platforms, including OpenClaw, for users with Claude subscriptions. According to Boris Cherny, head of Claude Code, users wishing to integrate such tools must now purchase discounted "extra usage bundles" linked to their Claude login or utilize a separate Claude API key from Anthropic's developer platform.

Anthropic attributed the decision to high compute demand from users. A company spokesperson stated that using Claude subscriptions with third-party tools violates the company's terms of service and places an "outsized strain on our systems." Cherny noted that the subscription model was not designed to accommodate the usage patterns of third-party tools and that capacity is managed with priority given to customers using Anthropic's direct products and API.

OpenClaw's Response

Peter Steinberger, creator of OpenClaw, and Dave Morin, a board member of the OpenClaw foundation, stated that they engaged with Anthropic to discuss the policy change. Steinberger reported that a significant number of users subscribed to Claude specifically for its integration with OpenClaw. The platform functions as an AI agent tool that connects users to models like Claude for various applications and workflows.

Industry Context

Anthropic's action follows a similar move by Google, which recently implemented restrictions on third-party tools for Gemini CLI users, citing terms of service violations.

App Store Performance

As of Saturday afternoon, Anthropic's Claude chatbot was ranked second among free applications in Apple's US App Store. According to SensorTower data, Claude was outside the top 100 at the end of January and remained largely within the top 20 throughout February. Its ranking rose from sixth on Wednesday to fourth on Thursday, reaching second place on Saturday. OpenAI's ChatGPT held the number one position, while Google Gemini was third.

Pentagon Negotiations and Government Response

Anthropic had engaged in negotiations with the Pentagon to implement safeguards aimed at preventing the Department of Defense from using its AI models for mass domestic surveillance or fully autonomous weapons.

Following these discussions, President Donald Trump issued a directive for federal agencies to cease using all Anthropic products. Secretary of Defense Pete Hegseth designated Anthropic as a supply-chain threat.

In a related development, OpenAI announced its own agreement with the Pentagon. CEO Sam Altman stated that this agreement includes safeguards related to domestic surveillance and autonomous weapons.

Additional Product Updates

Anthropic announced an update to Claude Design on Wednesday. The update introduces an administrator role that can set up a "design system" (a brand kit applicable across projects), with Claude automatically checking for compliance. Users can import GitHub repositories, design files, and raw uploads into the design system. Claude Code can now pull from designs and vice versa, using the command "/design."

Anthropic also pulled its newest AI model, Fable 5, after the US government ordered the company to prevent non-US citizens from using it. The export control order was issued following the discovery of a "jailbreak" that bypassed the model's cybersecurity guardrails. Anthropic stated that the only way to comply was to pull the model for all users.